Implementing a quantitative phenotypic covariate

[JC]

Perhaps this is question has an obvious answer that I'm just not finding, but regarding the addition of covariates, do the covariates have to be qualitative (like Cross as in the previous post) or can I use a quantitative covariate as well? I am mapping several phenotypes at once that are all correlated with biomass, and I'd like to include biomass as a covariate when I implement scanone(). I'm just not sure how to indicate to R/QTL that I want the biomass "phenotype" used as a covariate rather than mapped as a phenotype directly. Perhaps there's an argument I'm not familiar with? I've looked through the book and on the net, but all I seem to come up with is the use of marker covariates (as in CIM)...which isn't what I'm looking for.

I've considered regressing my phenotypes on biomass and then modifying the means by their respective regression coefficients to account for the effect, but I thought perhaps there is a more elegant way to do the same thing in R/QTL.

[Karl Broman]

You can include a quantitative covariate in QTL scans with no problem, and actually the covariates are assumed to be quantitative; with qualitative ones you have to ensure that you do the proper encoding (from a categorical variable to a set of binary variables).

Use care when including a secondary phenotype as a covariate. It changes the meaning of a QTL having an effect, and in fact QTL for biomass may show up as QTL for those other phenotypes even if there's no direct effect on those other phenotypes.

Code-wide, if I wanted to use the second phenotype as a covariate in analysis of the first, I'd do something like this:

mycross <- calc.genoprob(mycross, step=1)
covar <- mycross$pheno[,2]
out <- scanone(mycross, addcovar=covar)

karl

On May 23, 2012, at 9:48 AM, JC wrote:

> Perhaps this is question has an obvious answer that I'm just not finding, but regarding the addition of covariates, do the covariates have to be qualitative (like Cross as in the previous post) or can I use a quantitative covariate as well? I am mapping several phenotypes at once that are all correlated with biomass, and I'd like to include biomass as a covariate when I implement scanone(). I'm just not sure how to indicate to R/QTL that I want the biomass "phenotype" used as a covariate rather than mapped as a phenotype directly. Perhaps there's an argument I'm not familiar with? I've looked through the book and on the net, but all I seem to come up with is the use of marker covariates (as in CIM)...which isn't what I'm looking for.
>
> I've considered regressing my phenotypes on biomass and then modifying the means by their respective regression coefficients to account for the effect, but I thought perhaps there is a more elegant way to do the same thing in R/QTL.
>

> --
> You received this message because you are subscribed to the Google Groups "R/qtl discussion" group.
> To view this discussion on the web visit https://groups.google.com/d/msg/rqtl-disc/-/aoaxggdi4i4J.
> To post to this group, send email to rqtl...@googlegroups.com.
> To unsubscribe from this group, send email to rqtl-disc+...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/rqtl-disc?hl=en.